Diarization

In this branch, a new functionality has been added to the GMM and IVector algorithms, now you can extract i-vectors segment-wise and not only one i-vector per audio file.

bob/bio/gmm/algorithm/GMM.py

@@ -81,7 +81,7 @@ class GMM (Algorithm):
     self.scoring_function = scoring_function
 
     self.ubm = None
-    self.kmeans_trainer = bob.learn.em.KMeansTrainer()
+    self.kmeans_trainer = bob.learn.em.KMeansTrainer('RANDOM_NO_DUPLICATE')
     self.ubm_trainer = bob.learn.em.ML_GMMTrainer(self.update_means, self.update_variances, self.update_weights, self.responsibility_threshold)
 
 
@@ -155,6 +155,8 @@ class GMM (Algorithm):
 
   def load_ubm(self, ubm_file):
     hdf5file = bob.io.base.HDF5File(ubm_file)
+    if "/Projector" in hdf5file.sub_groups():
+      hdf5file.cd("/Projector")
     # read UBM
     self.ubm = bob.learn.em.GMMMachine(hdf5file)
     self.ubm.set_variance_thresholds(self.variance_threshold)
@@ -175,7 +177,6 @@ class GMM (Algorithm):
     # Accumulates statistics
     gmm_stats = bob.learn.em.GMMStats(self.ubm.shape[0], self.ubm.shape[1])
     self.ubm.acc_statistics(array, gmm_stats)
-
     # return the resulting statistics
     return gmm_stats